Metraipolis › archive › dc_dd5557fd5ce6
Paritok-4B: Intent-Conditioned Context Compression for Coding Agents — arXiv:2608.24188
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.24188
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T06:20:49.582Z
- SHA-256 of the archived text
f9351be7fd462845954124a2f97ddc7243cf8bb9b2ca55745a6869e2f59b3c26- Stable URL of this page
https://metraipolis.com/documento/dc_dd5557fd5ce6
How to cite
Paritok-4B: Intent-Conditioned Context Compression for Coding Agents — arXiv:2608.24188 — arXiv (en), https://arxiv.org/abs/2608.24188 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dd5557fd5ce6 (Metraipolis archive dc_dd5557fd5ce6, sha256 f9351be7fd462845954124a2f97ddc7243cf8bb9b2ca55745a6869e2f59b3c26)
BibTeX
@misc{dcdd5557fd5ce6,
title = {Paritok-4B: Intent-Conditioned Context Compression for Coding Agents — arXiv:2608.24188},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.24188},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dd5557fd5ce6; sha256 f9351be7fd462845954124a2f97ddc7243cf8bb9b2ca55745a6869e2f59b3c26},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.24188. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Coding agents re-send large file reads and tool outputs to a frontier LLM every turn, and this context dominates their token bill. General-purpose prompt compressors are trained on prose and suit code poorly: they paraphrase identifiers and drop the exact spans an agent needs to edit. We present Paritok-4B, a 4B LoRA compressor for coding-agent trajectories built on two commitments. It is extractive: it selects spans rather than rewriting them, and 96.0% of the identifiers, paths, and numbers it emits already appear in its input, holding at 96.2% on held-out SWE-bench Lite output. It is intent-conditioned: told the agent's current task, it acts chiefly inside a retained segment, selecting which lines survive (retained lines are +0.067 more intent-relevant than removed ones, paired 95% CI [+0.056, +0.078]) rather than changing how much is retained. We distil a gpt-4.1-mini teacher over 67,074 real OpenHands trajectories into 40,606 validated examples and fine-tune Qwen3-4B. On all 300 SWE-bench Lite instances, Paritok-4B compresses agent context to 25.7% of its size, 2.0x harder than a gpt-4.1-mini compressor (50.2%) and 2.4x harder than gpt-5 (61.9%), while retaining 86.5% of uncompressed single-shot solve quality. Fed the cat -n line-numbered input real agents produce, it compresses slightly less (27.8%) and retains more (89.3%); there the paired test is informative, with 30 instances solved only uncompressed and 17 only compressed, an exact McNemar p=0.079, so at this sample size compressing context to roughly a quarter of its size does not significantly reduce the solve rate. The model is a 264 MB adapter that self-hosts on one 24 GB GPU with no per-token compressor fee, which at list prices decides the economics: gpt-5 as a compressor is net-negative, costing more than the downstream tokens it saves. Weights, data, and evaluation scripts are open (Apache 2.0). Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-26 in cs.CL). Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Machine Learning (cs.LG); Software Engineering (cs.SE). Pagina del lavoro: https://arxiv.org/abs/2608.24188 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-18T06:17:29.405Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-26?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).